Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Multiscale full convolutional network feature fusion-based crowd counting method

A fully convolutional network and crowd counting technology, applied in biological neural network models, calculations, computer components, etc., can solve problems such as irregular distribution of crowds, achieve strong practicability, good robustness, and overcome occlusion effects

Pending Publication Date: 2018-09-28
SHANGHAI UNIV OF ENG SCI
View PDF4 Cites 27 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0008] In short, due to problems such as occlusion and irregular distribution of crowds, counting dense crowds is still facing great challenges. Therefore, developing a practical, robust, and accurate method for crowd counting or crowd density estimation will be of great help. It is of great significance to monitor the abnormal events of the crowd

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Multiscale full convolutional network feature fusion-based crowd counting method
  • Multiscale full convolutional network feature fusion-based crowd counting method
  • Multiscale full convolutional network feature fusion-based crowd counting method

Examples

Experimental program
Comparison scheme
Effect test

Embodiment

[0039] A crowd counting method based on multi-scale fully convolutional network feature fusion provided in this embodiment includes the following steps:

[0040] S1: Input a picture and enter three branch networks respectively to obtain feature maps of different scales; the specific operations are:

[0041] a) First convert the picture with the head position label of the person into a crowd density map, if there is a head position at the pixel point x i , denoting it as δ(x-x i ), then the image with the head position markers of N people can be expressed as the functional formula (1):

[0042]

[0043] b) Combine the functional formula (1) with the Gaussian kernel G σ Perform convolution to obtain the density estimation function (2):

[0044] F(x)=H(x)*G σ (x) (2)

[0045] c) Automatically determine the propagation parameter σ of each person based on the average distance data from its neighbors, if the distance from each head in a given image to its k nearest neighbors...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention discloses a multiscale full convolutional network feature fusion-based crowd counting method. The method comprises the following steps of: inputting a picture and respectively entering three branch networks so as to obtain feature maps with different scales; fusing the obtained feature maps of the branch networks so as to obtain a finally estimated feature map; mapping the output feature map into a density map; and carrying out summation on the density map to realize estimation of the current crowd number. The multiscale full convolutional network feature fusion-based crowd counting method provided by the invention is capable of overcoming severe influences, on crowd counting, of shielding, scene perspective distortion and different crowd distributions, has relatively strongpracticability and good robustness, is capable of accurately carrying out crowd counting or crowd density estimation, and has an important value for monitoring crowd abnormal events.

Description

technical field [0001] The invention relates to a crowd counting method based on multi-scale fully convolutional network feature fusion, and belongs to the technical field of image processing. Background technique [0002] Estimating people in dense settings has many potential practical applications, including surveillance (e.g., detecting unusually large crowds, or controlling the number of people in an area), security management (recording the number of people entering or leaving an area) , urban planning (for example, analyzing the flow of people in a certain area), etc. [0003] The methods for crowd counting in the prior art mainly include: [0004] 1) Pedestrian detection method: In the scene where the crowd is sparsely distributed, count by detecting each pedestrian in the video. This method is relatively straightforward, but suffers when occluded by crowds. [0005] 2) Trajectory clustering method: For surveillance videos, the KLT (Kanade-Lucas-Tomasi) tracker and...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Applications(China)
IPC IPC(8): G06K9/00G06K9/62G06N3/04
CPCG06V20/53G06N3/045G06F18/253
Inventor 方志军彭山珍高永彬黄勃韦钰
Owner SHANGHAI UNIV OF ENG SCI
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com